Experimental-ML frameworks for thermo-mechanical characterization of rubberized cement mortar after elevated temperature exposure

• 150 specimens were tested in compression and flexure to examine the thermo-mechanical behavior of rubberized cement mortar • DIC and ML models are employed to understand the influence of temperature on mixing proportions • Slow-cooled specimens recovered up to 20.1% more flexural strength than that obtained for instant-shock • Post-peak residual strength is enhanced by CR inclusion • Failure pattern transitioned from brittle to ductile in CR-incorporated mortar. This study presents a comprehensive experimental and data-driven investigation involving 150 specimens to evaluate the post-high-temperature thermo-mechanical behavior of rubberized cement mortar through compressive and flexural strength testing. Five mortar mixes are prepared, including a control (0% CR) and four mixes with 5%, 10%, 15%, and 20% CR by replacing NFA. Specimens are subjected to room temperature, 100°C, and 200°C, followed by immediate testing after thermal shock or slow cooling, enabling a systematic assessment of residual performance in high-temperature behavior. The findings of this investigation reveal that the compressive strength of rubberized mortar cubes decreased significantly with increasing CR content, with the highest reduction at 20% CR. At elevated temperatures, slow-cooled specimens recovered up to 20.1% more flexural strength than instant-shock counterparts, owing to improved stress redistribution and microcrack mitigation for 20% CR content. Notably, 5% CR replacement demonstrates comparable strength to the control mix while achieving superior strain accommodation and post-peak residual strength. Therefore, Digital Image Correlation (DIC) and Scanning Electron Microscopy (SEM) analyses are conducted on 5% CR beam specimens to quantify full-field strain evolution, crack development and microstructural damage mechanisms to further reinforce the experimental findings. Moreover, the current study presents the first machine learning-based approach toward estimating the flexural strength of rubberized cement mortar together with the impact level of temperature exposure, the post-heating testing environment, and the amount of CR%, with the results confirming that CR% influence is more prominent than the temperature on the flexural strength of rubberized mortar. However, this integrated experimental–DIC–SEM–ML framework represents a significant advancement over existing studies and may serve as a foundation for the optimized use of CR in thermally-exposed structural applications.